Ricci Flow-Based Approach for Early Diagnosis of Alzheimer's Disease

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Geometry

Background:

• Early diagnosis of Alzheimer's disease (AD) is crucial due to its increasing prevalence and impact.
• Hippocampal atrophy is a key biomarker for AD, detectable via MRI.
• Advanced imaging and surface analysis techniques show promise for faster, more accurate AD diagnosis.

Purpose of the Study:

• To apply the Ricci flow method for mapping 3D hippocampal surfaces to a 2D sphere.
• To extract features for early AD detection using Linear Discriminant Analysis (LDA) and Kernel LDA.
• To evaluate the diagnostic performance of the developed model using the ADNI dataset.

Main Methods:

• MRI scans were preprocessed to isolate hippocampal surfaces.
• Ricci flow was used to map the 3D hippocampal surface to a 2D sphere.
• Feature vectors were constructed using LDA and Kernel LDA, followed by classification and model evaluation.

Main Results:

• Combining Ricci flow features with Kernel LDA significantly enhanced diagnostic accuracy for AD detection.
• High classification accuracies were achieved: 97.28% (NC/AD), 96.14% (NC/EMCI), 96.45% (NC/MCI), 94.83% (EMCI/LMCI), 95.84% (MCI/AD), and 95.37% (LMCI/AD).
• Three-way and four-way classification tasks yielded accuracies of 93.65% and 92.30%, respectively, outperforming many existing studies.

Conclusions:

• The integration of Ricci flow-based feature extraction with Kernel LDA offers a powerful approach for precise AD diagnosis.
• This method demonstrates significant potential for improving early detection of Alzheimer's disease and its prodromal stages.
• The findings underscore the value of advanced imaging and mathematical modeling in enhancing diagnostic precision for neurodegenerative diseases.